consul/agent/structs/intention.go

741 lines
22 KiB
Go
Raw Normal View History

package structs
import (
"encoding/binary"
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
"encoding/json"
"errors"
2018-03-03 17:43:37 +00:00
"fmt"
"sort"
"strconv"
2018-03-03 17:43:37 +00:00
"strings"
"time"
2018-03-03 17:43:37 +00:00
"github.com/hashicorp/go-multierror"
"github.com/mitchellh/hashstructure"
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/cache"
"github.com/hashicorp/consul/lib"
"golang.org/x/crypto/blake2b"
)
const (
// IntentionDefaultNamespace is the default namespace value.
// NOTE(mitchellh): This is only meant to be a temporary constant.
// When namespaces are introduced, we should delete this constant and
// fix up all the places where this was used with the proper namespace
// value.
IntentionDefaultNamespace = "default"
)
// Intention defines an intention for the Connect Service Graph. This defines
// the allowed or denied behavior of a connection between two services using
// Connect.
type Intention struct {
// ID is the UUID-based ID for the intention, always generated by Consul.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
ID string `json:",omitempty"`
// Description is a human-friendly description of this intention.
// It is opaque to Consul and is only stored and transferred in API
// requests.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
Description string `json:",omitempty"`
// SourceNS, SourceName are the namespace and name, respectively, of
// the source service. Either of these may be the wildcard "*", but only
// the full value can be a wildcard. Partial wildcards are not allowed.
// The source may also be a non-Consul service, as specified by SourceType.
//
// DestinationNS, DestinationName is the same, but for the destination
// service. The same rules apply. The destination is always a Consul
// service.
SourceNS, SourceName string
DestinationNS, DestinationName string
// SourceType is the type of the value for the source.
SourceType IntentionSourceType
// Action is whether this is an allowlist or denylist intention.
Action IntentionAction `json:",omitempty"`
// Permissions is the list of additional L7 attributes that extend the
// intention definition.
//
// NOTE: This field is not editable unless editing the underlying
// service-intentions config entry directly.
Permissions []*IntentionPermission `bexpr:"-" json:",omitempty"`
// DefaultAddr is not used.
// Deprecated: DefaultAddr is not used and may be removed in a future version.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
DefaultAddr string `bexpr:"-" codec:",omitempty" json:",omitempty"`
// DefaultPort is not used.
// Deprecated: DefaultPort is not used and may be removed in a future version.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
DefaultPort int `bexpr:"-" codec:",omitempty" json:",omitempty"`
// Meta is arbitrary metadata associated with the intention. This is
// opaque to Consul but is served in API responses.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
Meta map[string]string `json:",omitempty"`
// Precedence is the order that the intention will be applied, with
// larger numbers being applied first. This is a read-only field, on
// any intention update it is updated.
Precedence int
// CreatedAt and UpdatedAt keep track of when this record was created
// or modified.
CreatedAt, UpdatedAt time.Time `mapstructure:"-" bexpr:"-"`
// Hash of the contents of the intention. This is only necessary for legacy
// intention replication purposes.
//
// This is needed mainly for legacy replication purposes. When replicating
// from one DC to another keeping the content Hash will allow us to detect
// content changes more efficiently than checking every single field
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
Hash []byte `bexpr:"-" json:",omitempty"`
RaftIndex `bexpr:"-"`
}
func (t *Intention) Clone() *Intention {
t2 := *t
if len(t.Permissions) > 0 {
t2.Permissions = make([]*IntentionPermission, 0, len(t.Permissions))
for _, perm := range t.Permissions {
t2.Permissions = append(t2.Permissions, perm.Clone())
}
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
t2.Meta = cloneStringStringMap(t.Meta)
t2.Hash = nil
return &t2
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
func (t *Intention) ToExact() *IntentionQueryExact {
return &IntentionQueryExact{
SourceNS: t.SourceNS,
SourceName: t.SourceName,
DestinationNS: t.DestinationNS,
DestinationName: t.DestinationName,
}
}
func (t *Intention) MarshalJSON() ([]byte, error) {
type Alias Intention
exported := &struct {
CreatedAt, UpdatedAt *time.Time `json:",omitempty"`
*Alias
}{
Alias: (*Alias)(t),
}
if !t.CreatedAt.IsZero() {
exported.CreatedAt = &t.CreatedAt
}
if !t.UpdatedAt.IsZero() {
exported.UpdatedAt = &t.UpdatedAt
}
return json.Marshal(exported)
}
func (t *Intention) UnmarshalJSON(data []byte) (err error) {
type Alias Intention
aux := &struct {
Hash string
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
CreatedAt, UpdatedAt string // effectively `json:"-"` on CreatedAt and UpdatedAt
*Alias
}{
Alias: (*Alias)(t),
}
if err = lib.UnmarshalJSON(data, &aux); err != nil {
return err
}
if aux.Hash != "" {
t.Hash = []byte(aux.Hash)
}
return nil
}
// SetHash calculates Intention.Hash from any mutable "content" fields.
//
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// The Hash is primarily used for legacy intention replication to determine if
// an intention has changed and should be updated locally.
//
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// Deprecated: this is only used for legacy intention CRUD and replication
func (x *Intention) SetHash() {
hash, err := blake2b.New256(nil)
if err != nil {
panic(err)
}
// Write all the user set fields
hash.Write([]byte(x.ID))
hash.Write([]byte(x.Description))
hash.Write([]byte(x.SourceNS))
hash.Write([]byte(x.SourceName))
hash.Write([]byte(x.DestinationNS))
hash.Write([]byte(x.DestinationName))
hash.Write([]byte(x.SourceType))
hash.Write([]byte(x.Action))
// hash.Write can not return an error, so the only way for binary.Write to
// error is to pass it data with an invalid data type. Doing so would be a
// programming error, so panic in that case.
if err := binary.Write(hash, binary.LittleEndian, uint64(x.Precedence)); err != nil {
panic(err)
}
// sort keys to ensure hash stability when meta is stored later
var keys []string
for k := range x.Meta {
keys = append(keys, k)
}
sort.Strings(keys)
for _, k := range keys {
hash.Write([]byte(k))
hash.Write([]byte(x.Meta[k]))
}
x.Hash = hash.Sum(nil)
}
2018-03-03 17:43:37 +00:00
// Validate returns an error if the intention is invalid for inserting
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// or updating via the legacy APIs.
//
// Deprecated: this is only used for legacy intention CRUD
2018-03-03 17:43:37 +00:00
func (x *Intention) Validate() error {
var result error
// Empty values
if x.SourceNS == "" {
result = multierror.Append(result, fmt.Errorf("SourceNS must be set"))
}
if x.SourceName == "" {
result = multierror.Append(result, fmt.Errorf("SourceName must be set"))
}
if x.DestinationNS == "" {
result = multierror.Append(result, fmt.Errorf("DestinationNS must be set"))
}
if x.DestinationName == "" {
result = multierror.Append(result, fmt.Errorf("DestinationName must be set"))
}
// Wildcard usage verification
if x.SourceNS != WildcardSpecifier {
if strings.Contains(x.SourceNS, WildcardSpecifier) {
2018-03-03 17:43:37 +00:00
result = multierror.Append(result, fmt.Errorf(
"SourceNS: wildcard character '*' cannot be used with partial values"))
}
}
if x.SourceName != WildcardSpecifier {
if strings.Contains(x.SourceName, WildcardSpecifier) {
2018-03-03 17:43:37 +00:00
result = multierror.Append(result, fmt.Errorf(
"SourceName: wildcard character '*' cannot be used with partial values"))
}
if x.SourceNS == WildcardSpecifier {
2018-03-03 17:43:37 +00:00
result = multierror.Append(result, fmt.Errorf(
"SourceName: exact value cannot follow wildcard namespace"))
}
}
if x.DestinationNS != WildcardSpecifier {
if strings.Contains(x.DestinationNS, WildcardSpecifier) {
2018-03-03 17:43:37 +00:00
result = multierror.Append(result, fmt.Errorf(
"DestinationNS: wildcard character '*' cannot be used with partial values"))
}
}
if x.DestinationName != WildcardSpecifier {
if strings.Contains(x.DestinationName, WildcardSpecifier) {
2018-03-03 17:43:37 +00:00
result = multierror.Append(result, fmt.Errorf(
"DestinationName: wildcard character '*' cannot be used with partial values"))
}
if x.DestinationNS == WildcardSpecifier {
2018-03-03 17:43:37 +00:00
result = multierror.Append(result, fmt.Errorf(
"DestinationName: exact value cannot follow wildcard namespace"))
}
}
// Length of opaque values
if len(x.Description) > metaValueMaxLength {
result = multierror.Append(result, fmt.Errorf(
"Description exceeds maximum length %d", metaValueMaxLength))
}
if len(x.Meta) > metaMaxKeyPairs {
result = multierror.Append(result, fmt.Errorf(
"Meta exceeds maximum element count %d", metaMaxKeyPairs))
}
for k, v := range x.Meta {
if len(k) > metaKeyMaxLength {
result = multierror.Append(result, fmt.Errorf(
"Meta key %q exceeds maximum length %d", k, metaKeyMaxLength))
}
if len(v) > metaValueMaxLength {
result = multierror.Append(result, fmt.Errorf(
"Meta value for key %q exceeds maximum length %d", k, metaValueMaxLength))
}
}
switch x.Action {
case IntentionActionAllow, IntentionActionDeny:
default:
result = multierror.Append(result, fmt.Errorf(
"Action must be set to 'allow' or 'deny'"))
}
if len(x.Permissions) > 0 {
result = multierror.Append(result, fmt.Errorf(
"Permissions must not be set when using the legacy APIs"))
}
switch x.SourceType {
case IntentionSourceConsul:
default:
result = multierror.Append(result, fmt.Errorf(
"SourceType must be set to 'consul'"))
}
2018-03-03 17:43:37 +00:00
return result
}
func (ixn *Intention) CanRead(authz acl.Authorizer) bool {
var authzContext acl.AuthorizerContext
// Read access on either end of the intention allows you to read the
// complete intention. This is so that both ends can be aware of why
// something does or does not work.
if ixn.SourceName != "" {
ixn.FillAuthzContext(&authzContext, false)
if authz.IntentionRead(ixn.SourceName, &authzContext) == acl.Allow {
return true
}
}
if ixn.DestinationName != "" {
ixn.FillAuthzContext(&authzContext, true)
if authz.IntentionRead(ixn.DestinationName, &authzContext) == acl.Allow {
return true
}
}
return false
}
func (ixn *Intention) CanWrite(authz acl.Authorizer) bool {
var authzContext acl.AuthorizerContext
ixn.FillAuthzContext(&authzContext, true)
return authz.IntentionWrite(ixn.DestinationName, &authzContext) == acl.Allow
}
// UpdatePrecedence sets the Precedence value based on the fields of this
// structure.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
//
// Deprecated: this is only used for legacy intention CRUD.
func (x *Intention) UpdatePrecedence() {
// Max maintains the maximum value that the precedence can be depending
// on the number of exact values in the destination.
var max int
switch x.countExact(x.DestinationNS, x.DestinationName) {
case 2:
max = 9
case 1:
max = 6
case 0:
max = 3
default:
// This shouldn't be possible, just set it to zero
x.Precedence = 0
return
}
// Given the maximum, the exact value is determined based on the
// number of source exact values.
countSrc := x.countExact(x.SourceNS, x.SourceName)
x.Precedence = max - (2 - countSrc)
}
// countExact counts the number of exact values (not wildcards) in
// the given namespace and name.
func (x *Intention) countExact(ns, n string) int {
// If NS is wildcard, it must be zero since wildcards only follow exact
if ns == WildcardSpecifier {
return 0
}
// Same reasoning as above, a wildcard can only follow an exact value
// and an exact value cannot follow a wildcard, so if name is a wildcard
// we must have exactly one.
if n == WildcardSpecifier {
return 1
}
return 2
}
// String returns a human-friendly string for this intention.
func (x *Intention) String() string {
var idPart string
if x.ID != "" {
idPart = "ID: " + x.ID + ", "
}
var detailPart string
if len(x.Permissions) > 0 {
detailPart = fmt.Sprintf("Permissions: %d", len(x.Permissions))
} else {
detailPart = "Action: " + strings.ToUpper(string(x.Action))
}
return fmt.Sprintf("%s/%s => %s/%s (%sPrecedence: %d, %s)",
x.SourceNS, x.SourceName,
x.DestinationNS, x.DestinationName,
idPart,
x.Precedence,
detailPart,
)
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// LegacyEstimateSize returns an estimate (in bytes) of the size of this structure when encoded.
//
// Deprecated: only exists for legacy intention replication during migration to 1.9.0+ cluster.
func (x *Intention) LegacyEstimateSize() int {
// 56 = 36 (uuid) + 16 (RaftIndex) + 4 (Precedence)
size := 56 + len(x.Description) + len(x.SourceNS) + len(x.SourceName) + len(x.DestinationNS) +
len(x.DestinationName) + len(x.SourceType) + len(x.Action)
for k, v := range x.Meta {
size += len(k) + len(v)
}
return size
}
func (x *Intention) SourceServiceName() ServiceName {
return NewServiceName(x.SourceName, x.SourceEnterpriseMeta())
}
func (x *Intention) DestinationServiceName() ServiceName {
return NewServiceName(x.DestinationName, x.DestinationEnterpriseMeta())
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// NOTE this is just used to manipulate user-provided data before an insert
// The RPC execution will do Normalize + Validate for us.
func (x *Intention) ToConfigEntry(legacy bool) *ServiceIntentionsConfigEntry {
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
return &ServiceIntentionsConfigEntry{
Kind: ServiceIntentions,
Name: x.DestinationName,
EnterpriseMeta: *x.DestinationEnterpriseMeta(),
Sources: []*SourceIntention{x.ToSourceIntention(legacy)},
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
}
func (x *Intention) ToSourceIntention(legacy bool) *SourceIntention {
ct := x.CreatedAt // copy
ut := x.UpdatedAt
src := &SourceIntention{
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
Name: x.SourceName,
EnterpriseMeta: *x.SourceEnterpriseMeta(),
Action: x.Action,
Permissions: nil, // explicitly not symmetric with the old APIs
Precedence: 0, // Ignore, let it be computed.
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
LegacyID: x.ID,
Type: x.SourceType,
Description: x.Description,
LegacyMeta: x.Meta,
LegacyCreateTime: &ct,
LegacyUpdateTime: &ut,
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
if !legacy {
src.Permissions = x.Permissions
}
return src
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
}
// IntentionAction is the action that the intention represents. This
// can be "allow" or "deny".
type IntentionAction string
const (
IntentionActionAllow IntentionAction = "allow"
IntentionActionDeny IntentionAction = "deny"
)
// IntentionSourceType is the type of the source within an intention.
type IntentionSourceType string
const (
// IntentionSourceConsul is a service within the Consul catalog.
IntentionSourceConsul IntentionSourceType = "consul"
)
// Intentions is a list of intentions.
type Intentions []*Intention
// IndexedIntentions represents a list of intentions for RPC responses.
type IndexedIntentions struct {
Intentions Intentions
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
// DataOrigin is used to indicate if this query was satisfied against the
// old legacy intentions ("legacy") memdb table or via config entries
// ("config"). This is really only of value for the legacy intention
// replication routine to correctly detect that it should exit.
DataOrigin string `json:"-"`
QueryMeta
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
const (
IntentionDataOriginLegacy = "legacy"
IntentionDataOriginConfigEntries = "config"
)
// IndexedIntentionMatches represents the list of matches for a match query.
type IndexedIntentionMatches struct {
Matches []Intentions
QueryMeta
}
// IntentionOp is the operation for a request related to intentions.
type IntentionOp string
const (
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
IntentionOpCreate IntentionOp = "create"
IntentionOpUpdate IntentionOp = "update"
IntentionOpDelete IntentionOp = "delete"
IntentionOpDeleteAll IntentionOp = "delete-all" // NOTE: this is only accepted when it comes from the leader, RPCs will reject this
IntentionOpUpsert IntentionOp = "upsert" // config-entry only
)
// IntentionRequest is used to create, update, and delete intentions.
type IntentionRequest struct {
// Datacenter is the target for this request.
Datacenter string
// Op is the type of operation being requested.
Op IntentionOp
// Intention is the intention.
//
// This is mutually exclusive with the Mutation field.
Intention *Intention
// Mutation is a change to make to an Intention.
//
// This is mutually exclusive with the Intention field.
//
// This field is only set by the leader before writing to the raft log and
// is not settable via the API or an RPC.
Mutation *IntentionMutation
// WriteRequest is a common struct containing ACL tokens and other
// write-related common elements for requests.
WriteRequest
}
type IntentionMutation struct {
ID string
Destination ServiceName
Source ServiceName
Value *SourceIntention
}
// RequestDatacenter returns the datacenter for a given request.
func (q *IntentionRequest) RequestDatacenter() string {
return q.Datacenter
}
2018-02-28 18:44:49 +00:00
// IntentionMatchType is the target for a match request. For example,
// matching by source will look for all intentions that match the given
// source value.
type IntentionMatchType string
const (
IntentionMatchSource IntentionMatchType = "source"
IntentionMatchDestination IntentionMatchType = "destination"
)
2018-02-28 18:44:49 +00:00
// IntentionQueryRequest is used to query intentions.
type IntentionQueryRequest struct {
// Datacenter is the target this request is intended for.
Datacenter string
// IntentionID is the ID of a specific intention.
IntentionID string
// Match is non-nil if we're performing a match query. A match will
// find intentions that "match" the given parameters. A match includes
// resolving wildcards.
Match *IntentionQueryMatch
2018-05-11 16:19:22 +00:00
// Check is non-nil if we're performing a test query. A test will
// return allowed/deny based on an exact match.
2018-05-11 16:19:22 +00:00
Check *IntentionQueryCheck
// Exact is non-nil if we're performing a lookup of an intention by its
// unique name instead of its ID.
Exact *IntentionQueryExact
2018-02-28 18:44:49 +00:00
// Options for queries
QueryOptions
}
// RequestDatacenter returns the datacenter for a given request.
func (q *IntentionQueryRequest) RequestDatacenter() string {
return q.Datacenter
}
// CacheInfo implements cache.Request
func (q *IntentionQueryRequest) CacheInfo() cache.RequestInfo {
info := cache.RequestInfo{
Token: q.Token,
Datacenter: q.Datacenter,
MinIndex: q.MinQueryIndex,
Timeout: q.MaxQueryTime,
}
v, err := hashstructure.Hash(struct {
IntentionID string
Match *IntentionQueryMatch
Check *IntentionQueryCheck
Exact *IntentionQueryExact
Filter string
}{
IntentionID: q.IntentionID,
Check: q.Check,
Match: q.Match,
Exact: q.Exact,
Filter: q.QueryOptions.Filter,
}, nil)
if err == nil {
// If there is an error, we don't set the key. A blank key forces
// no cache for this request so the request is forwarded directly
// to the server.
info.Key = strconv.FormatUint(v, 16)
}
return info
}
// IntentionQueryMatch are the parameters for performing a match request
// against the state store.
type IntentionQueryMatch struct {
Type IntentionMatchType
Entries []IntentionMatchEntry
}
// IntentionMatchEntry is a single entry for matching an intention.
type IntentionMatchEntry struct {
Namespace string
Name string
}
2018-05-11 16:19:22 +00:00
// IntentionQueryCheck are the parameters for performing a test request.
type IntentionQueryCheck struct {
// SourceNS, SourceName, DestinationNS, and DestinationName are the
// source and namespace, respectively, for the test. These must be
// exact values.
SourceNS, SourceName string
DestinationNS, DestinationName string
// SourceType is the type of the value for the source.
SourceType IntentionSourceType
}
// GetACLPrefix returns the prefix to look up the ACL policy for this
// request, and a boolean noting whether the prefix is valid to check
// or not. You must check the ok value before using the prefix.
2018-05-11 16:19:22 +00:00
func (q *IntentionQueryCheck) GetACLPrefix() (string, bool) {
return q.DestinationName, q.DestinationName != ""
}
2018-05-11 16:19:22 +00:00
// IntentionQueryCheckResponse is the response for a test request.
type IntentionQueryCheckResponse struct {
Allowed bool
}
// IntentionDecisionSummary contains a summary of a set of intentions between two services
// Currently contains:
// - Whether all actions are allowed
// - Whether the matching intention has L7 permissions attached
// - Whether the intention is managed by an external source like k8s
// - Whether there is an exact, or wildcard, intention referencing the two services
// - Whether ACLs are in DefaultAllow mode
type IntentionDecisionSummary struct {
Allowed bool
HasPermissions bool
ExternalSource string
HasExact bool
DefaultAllow bool
}
// IntentionQueryExact holds the parameters for performing a lookup of an
// intention by its unique name instead of its ID.
type IntentionQueryExact struct {
SourceNS, SourceName string
DestinationNS, DestinationName string
}
// Validate is used to ensure all 4 parameters are specified.
func (q *IntentionQueryExact) Validate() error {
var err error
if q.SourceNS == "" {
err = multierror.Append(err, errors.New("SourceNS is missing"))
}
if q.SourceName == "" {
err = multierror.Append(err, errors.New("SourceName is missing"))
}
if q.DestinationNS == "" {
err = multierror.Append(err, errors.New("DestinationNS is missing"))
}
if q.DestinationName == "" {
err = multierror.Append(err, errors.New("DestinationName is missing"))
}
return err
}
connect: intentions are now managed as a new config entry kind "service-intentions" (#8834) - Upgrade the ConfigEntry.ListAll RPC to be kind-aware so that older copies of consul will not see new config entries it doesn't understand replicate down. - Add shim conversion code so that the old API/CLI method of interacting with intentions will continue to work so long as none of these are edited via config entry endpoints. Almost all of the read-only APIs will continue to function indefinitely. - Add new APIs that operate on individual intentions without IDs so that the UI doesn't need to implement CAS operations. - Add a new serf feature flag indicating support for intentions-as-config-entries. - The old line-item intentions way of interacting with the state store will transparently flip between the legacy memdb table and the config entry representations so that readers will never see a hiccup during migration where the results are incomplete. It uses a piece of system metadata to control the flip. - The primary datacenter will begin migrating intentions into config entries on startup once all servers in the datacenter are on a version of Consul with the intentions-as-config-entries feature flag. When it is complete the old state store representations will be cleared. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up. - The secondary datacenters continue to run the old intentions replicator until all servers in the secondary DC and primary DC support intentions-as-config-entries (via serf flag). Once this condition it met the old intentions replicator ceases. - The secondary datacenters replicate the new config entries as they are migrated in the primary. When they detect that the primary has zeroed it's old state store table it waits until all config entries up to that point are replicated and then zeroes its own copy of the old state store table. We also record a piece of system metadata indicating this has occurred. We use this metadata to skip ALL of this code the next time the leader starts up.
2020-10-06 18:24:05 +00:00
type IntentionListRequest struct {
Datacenter string
Legacy bool `json:"-"`
EnterpriseMeta `hcl:",squash" mapstructure:",squash"`
QueryOptions
}
func (r *IntentionListRequest) RequestDatacenter() string {
return r.Datacenter
}
// IntentionPrecedenceSorter takes a list of intentions and sorts them
// based on the match precedence rules for intentions. The intentions
// closer to the head of the list have higher precedence. i.e. index 0 has
// the highest precedence.
type IntentionPrecedenceSorter Intentions
func (s IntentionPrecedenceSorter) Len() int { return len(s) }
func (s IntentionPrecedenceSorter) Swap(i, j int) {
s[i], s[j] = s[j], s[i]
}
func (s IntentionPrecedenceSorter) Less(i, j int) bool {
a, b := s[i], s[j]
if a.Precedence != b.Precedence {
return a.Precedence > b.Precedence
}
// Tie break on lexicographic order of the 4-tuple in canonical form (SrcNS,
// Src, DstNS, Dst). This is arbitrary but it keeps sorting deterministic
// which is a nice property for consistency. It is arguably open to abuse if
// implementations rely on this however by definition the order among
// same-precedence rules is arbitrary and doesn't affect whether an allow or
// deny rule is acted on since all applicable rules are checked.
if a.SourceNS != b.SourceNS {
return a.SourceNS < b.SourceNS
}
if a.SourceName != b.SourceName {
return a.SourceName < b.SourceName
}
if a.DestinationNS != b.DestinationNS {
return a.DestinationNS < b.DestinationNS
}
return a.DestinationName < b.DestinationName
}